Research in Business and Economics Journal, Volume 10 – October, 2014
Cost behavior of selling, general, and administrative costs and cost of
goods sold during economic recession
Haihong He
California State University, Los Angeles
ABSTRACT
Cost management creates value for firms. Effective cost management leads to a firm's success.
At difficult times, good cost management means survival. This paper gauges firms' cost management
practices by empirically examining cost behavior of selling, general and administrative costs and cost
of goods sold during the 2008 – 2009 recession.
This paper empirically investigates how selling, general and administrative costs and cost of
goods sold behave in the recession period with the pre-recession period as a benchmark. We find that,
although the total costs (as a percentage of sales revenue) on average do no differ, both SGA and
COGS changes became less sensitive to sales revenue changes in the recession period, and the
stickiness of the two costs also changed but in different directions in the recession period. Since costs
and activity relationship is greatly influenced by management decisions, we interpret our findings as
results of managers coping crises with different cost management practices.
Keywords: cost behavior, cost of good sold, administrative costs, recession
Copyright statement: Authors retain the copyright to the manuscripts published in AABRI journals.
Please see the AABRI Copyright Policy at http://www.aabri.com/copyright.html
Cost behavior of selling, Page 1
Research in Business and Economics Journal, Volume 10 – October, 2014
INTRODUCTION
Cost management creates value for firms. Effective cost management leads to a firm's success.
At difficult times, good cost management means survival. This paper gauges firms' cost management
practices by empirically examining cost behavior of selling, general and administrative costs and cost
of goods sold during the 2008 – 2009 recession.
Theoretically, costs display two different behaviors. Variable costs change proportionally in
response to the activity change while fixed costs remain constant. In essence, there is a linear
relationship between costs and activity (such as production units, number of customers) that cause the
costs to occur. Prior research ( Banker and Johnston 1993; Noreen and Soderstrom 1994, 1997) found
that, however, costs rarely follow this theoretical pattern strictly. They either show stickiness or antistickiness. Stickiness means that for equal increase or decrease in activity, costs increase is greater than
costs decrease. Thus, the costs change is not proportional to the activity change. Anderson et al. (2003)
attribute such findings to managers deliberately adjust firm resources in different circumstances.
We follow prior study (Anderson et al. 2003) to test costs variability with respect to sales
revenue variability during the recession period from 2008 – 2009 and compare to that of the pre recession period from 2005 – 2006. The different cost behavior in the two periods can provide us some
insight into the managerial practice shift which is reflected in the cost management. We study selling
and general administrative costs and cost of goods sold.
PRIOR RESEARCH AND RESEARCH QUESTIONS
Costs are generally assumed to be proportional to activity levels (Garrison et al. 2011).
However, empirical research (Banker and Johnston 1993; Noreen and Soderstrom 1994, 1997) do not
indicate such proportional relationship. A few recent research such as Anderson et al. (2003) further
show that some costs are sticky - costs increase more when activities rise than they decreases when
activities fall. Some studies (Balakrishnan et al. 2003) show that some costs are anti-sticky – costs
decrease more when activities decrease than they increase when activities rise.
Costs do not follow the theoretical model strictly because managerial action in response to
changed circumstances cause cost behavior deviate from the proportional and mechanical relationship
(Anderson et al. 2003). Anderson et al. (2003) offers resource adjustment costs as an explanation to the
stickiness: it is more costly to reduce resources when sales decreases than increase resources when
sales increases. Banker et al. (2014) add evidence to support the deliberate managerial decisions and
cost behavior. They show that cost stickiness is conditional on a prior sales increase and cost antistikiness is conditional on a prior sales decrease. Kama and Weiss (2013) shows that when managers
are incentivised to avoid losses or earnings decreases, they expedite downward adjustment of slack
resources for sales decreases, which is anti-stickiness. Kallapur and Eldenburg (2005) show that, in the
face of uncertainty, firms will prefer low fixed and high variable costs.
Based on the prior research, we study costs variability in the 2008 – 2009 recession period, and
compare it to the 2005 – 2006 period to show how managers make moves to cope with the macroeconomic and financial difficulties. Our research questions are as follows:
1. Does costs to sales revenue ratio differ between the two periods?
2. Does costs change in reaction to sales revenue change differ between the two periods?
3. Does costs stickiness differ between the two periods?
Cost behavior of selling, Page 2
Research in Business and Economics Journal, Volume 10 – October, 2014
SAMPLE
We obtain all U.S. firms from Compustat for two periods: pre-recession period 2005 – 2006 and
recession period 2008 – 2009. We do not include year 2007 because some U.S. firm started to face
some economic difficulties since 2007 and including them may cause our results hard to interpret. We
also delete financial institutions (one-digit SIC = 6) and utility firms (two-digit SIC = 49) because they
are in highly regulated industries and may have different cost management practice. In addition to cost
variables, selling, general, and administrative costs (mnemonics XSGA) and Cost of goods sold
(mnemonics XLR), we also obtained sales revenue (mnemonics SALE). We further delete any
observations with negative sales revenue. Finally we delete observations with sales revenue greater
than selling, general, and administrative costs and sales greater than cost of goods sold.
RESULTS
Table 1 reports costs as a percentage of sales in the two periods. After meeting data
requirements, final sample consists of 4,319 observations in the 2005 – 2006 period and 6,727
observations in the 2008 – 2009 period. Costs to sales ratio appear to be fairly similar in the two
periods. Selling, general, and administrative costs ratio (SGA) has a mean of 0.3015 and 0.2984 in the
two periods. Cost of goods sold ratio (COGS) has a mean of 0.5924 and 0.5960 in the two periods.
Overall, the two costs combined are about 89.40% of sales in the 2005 – 2006 period and 89.44% of
sales in the 2008 – 2009 period. Further, we perform a t-test (untabulated) which do not show any
statistical difference of these ratios across the two periods. See Table 1 in the appendix.
To address the second research question, we regress change in costs on change in sales revenue.
The change variables are computed as the logarithm of the ratio of current value to the previous period
value. The model used as follows is modified from Anderson et al. (2003):
log ( cost t / cost (t − 1 ))= β0+ β1∗log ( Revenue t / Revenue (t − 1))+ ϵ
We report OLS regression results in Table 2. β0 is the estimated coefficient on the intercept and
β1 is the estimated coefficient on change in sales revenue that measures the change in costs in reaction
to the changes in sales revenue. The results are consistent with SGA's mixed cost behavior and COGS's
variable cost behavior with sales revenue as the activity measure. For example, in Table 2 Panel A, β0
(the fixed component) is only significant in the SGA regression, but insignificant in the COGS
regression; the estimated β1 =0.5128 means that SGA changes about 0.51% when sales revenue changes
1%, and the estimated β1 =1.0126 means that COGS changes about 1% when sales revenue change
1%. Then we compare β1 across the two periods as reported in Panel A and Panel B. In the SGA
regression, β1 is 0.5128 in the 2005 – 2006 period and 0.4657 in the 2008 – 2009 period, and a t- test
show this difference is significant (t=3.86). The results suggest that SGA increase less with respect to
sales revenue increase in the latter period, which indicates greater cost control. In the COGS regression,
β1 is 1.0126 in the 2005 – 2006 period and 0.9047 in the 2008 – 2009 period, and a t- test show this
difference is also significant (t=10.86). There could be a few explanations for why COGS change less
proportionally with sales revenue in the latter period. First, our activity measured by sales revenue in
dollars instead of sales units contribute to such results. Second, related to using sales revenue as
activity measure, managers may change products mix in the latter period, and this could be a result of
managers focusing resources on high profit margin products in the difficult economies. See Table 2 in
the appendix.
To address the third research question, following Anderson et al. (2003), we regress change in
Cost behavior of selling, Page 3
Research in Business and Economics Journal, Volume 10 – October, 2014
costs on change in sales revenue and a dummy variable for sales revenue change. The change variables
are computed as the logarithm of the ratio of current value to the previous period value. The dummy
variable is 1 if sales revenue decreases. Results of estimating this model are reported in Table 3. The
model used is as follows:
log( cost t /cost (t − 1))= β0+ β1∗log( Revenue t / Revenue(t − 1 ))+ β2∗Dummy∗log( Revenue t / Revenue (t − 1))+ ϵ
Including a indicator variable, this model is slightly different than the model for Table 2, and
has been used extensively in studying stickiness of cost behavior. If β2 is significantly negative, then
costs decreases less than increases for the same magnitude of sales decrease and increase, a sticky cost
behavior. OLS regression results are reported in Table 3. In SGA regression, β2 is 0.1273 (t=4.73) in the
2005 – 2006 period and insignificant in the 2008 – 2009 period, and this difference is significant
(t=4.54). These suggest that SGA decrease more than increase to respond to sales change, but such
different reactions no longer exist in the 2008 – 2009 period. In COGS regression, β2 is insignificant in
the 2005 – 2006 period and 0.1251 and significant (t=6.89) in the 2008 – 2009 period, and this
difference is significant (t=3.23). It is clear that stickiness of SGA and COGS is different in the two
periods. See Table 3 in the appendix.
CONCLUSIONS
This paper empirically investigates how selling, general and administrative costs and cost of
goods sold behave in the recession period with the pre-recession period as a benchmark. We find that,
although the total costs (as a percentage of sales revenue) on average do no differ, both SGA and
COGS changes became less sensitive to sales revenue changes in the recession period, and the
stickiness of the two costs also changed but in different directions in the recession period. Since costs
and activity relationship is greatly influenced by management decisions, we interpret our findings as
results of managers coping crises with different cost management practices.
REFERENCE
Anderson, M. C., R. D. Banker, and S.N. Janakiraman. 2003. Are selling, general, and administrative
costs “sticky”? Journal of Accounting Research 41 (1): 47-63.
Balakrishnan, R., M. Peterson, and N. Soderstrom. 2003. Does capacity utilization affect the
“stickiness” of cost? Working Paper.
Banker, R. D., and H. H. Johnston. 1993. An empirical study of cost drivers in the U.S. Airline industry.
The Accounting Review 68: 576-601.
Banker, R. D., D. Byzalov, M. Ciftci, and R. Mashruwala. 2014. The moderating effect of prior sales
changes on asymmetric cost behavior. Journal of Management Accounting Research Forthcoming.
Garrison, R., E. Noreen and P. Brewer. 2014. Managerial Accounting, 14th edition.
Kallapur, S. and L. Eldenburg. 2005. Uncertainty, real options, and cost behavior: evidence from
Washington state hospitals. Journal of Accounting Research 43 (5): 735-752.
Cost behavior of selling, Page 4
Research in Business and Economics Journal, Volume 10 – October, 2014
Kama, I. And D. Weiss. 2013. Do earnings targets and managerial incentives affect sticky costs?
Journal of Accounting Research 2013 (March): 201-224.
Noreen, E. and N. Soderstrom. 1994. Are overhead costs strictly proportional to activity? Journal of
Accounting and Economics 17 (½): 255-278.
Noreen, E. and N. Soderstrom. 1997. The accuracy of proportional cost models: Evidence from
hospital service departments. Review of Accounting Studies 2: 89-114.
Subramaniam, C. and M . L. Weidenmier. 2003. Additional evidence on the sticky behavior of costs.
Working Paper.
Cost behavior of selling, Page 5
Research in Business and Economics Journal, Volume 10 – October, 2014
Table 1 Descriptive Statistics of Costs as a Percentage of Sales Revenue
Panel A 2005 – 2006
Mean
Std.
Median
Lower
Quartile
Upper
Quartile
N
SG&A costs as a
percentage of revenue
0.3015
0.2102
0.2564
0.1389
0.4098
4,319
COGS as a
percentage of revenue
0.5924
0.2077
0.6220
0.4497
0.7541
4,319
Total costs as a
percentage of revenue
0.8940
0.1730
0.9002
0.8215
0.9629
4,319
Mean
Std.
Median
Lower
Quartile
Upper
Quartile
N
SG&A costs as a
percentage of revenue
0.2984
0.2122
0.2493
0.1305
0.4148
6,727
COGS as a
percentage of revenue
0.5960
0.2110
0.6296
0.4487
0.7587
6,727
Total costs as a
percentage of revenue
0.8944
0.1728
0.9013
0.8141
0.9708
6,727
Panel B 2008 – 2009
Cost behavior of selling, Page 6
Research in Business and Economics Journal, Volume 10 – October, 2014
Table 2 Results of Regressing Changes in SG&A on Changes in Sales Revenue for 2005-2006 and
2008-2009 Periods (t statistics in pare
log ( cost t / cost (t − 1))= β0+ β1∗log( Revenue t / Revenue(t − 1 ))+ ϵ
Panel A 2005 – 2006 Period Parameter Estimates (t-statistics in parentheses)
SGA
COGS
β0
0.0578***
-0.0005
(16.47)
(0.17)
0.5128***
1.0126***
(58.87)
(144.05)
Adjusted R2
0.4452
0.8278
N
4,319
4,319
β1
Panel B 2008 – 2009 Period Parameter Estimates (t-statistics in parentheses)
SGA
COGS
β0
0.0153***
-0.0024
(5.46)
(1.01)
0.4657***
0.9047***
(55.71)
(132.35)
Adjusted R2
0.3157
0.7225
N
6,727
6,727
β1
***, **, * statistically significant at the 1%, 5%, 10% levels for a two-tailed test.
When β1 is compared across the two periods, it is significantly lower in the SGA regression (t=3.86) and
significantly lower in the COGS regression (t=10.86) in the 2008 – 2009 period.
Cost behavior of selling, Page 7
Research in Business and Economics Journal, Volume 10 – October, 2014
Table 3 Results of Regressing Changes in SG&A on Changes in Sales Revenue and Sales Decrease
Indicator for 2005-2006 and 2008-2009 Periods
log( cost t /cost (t − 1))= β0+ β1∗log( Revenue t / Revenue(t − 1 ))+ β2∗Dummy∗log( Revenue t / Revenue (t − 1))+ ϵ
Panel A 2005 – 2006 Period Parameter Estimates (t-statistics in parentheses)
SG&A
COGS
β0
0.0657***
0.0014
(16.97)
(0.46)
0.4899***
1.007***
(48.22)
(125.07)
0.1273**
0.0318
(4.73)
(1.46)
Adjusted R2
0.4479
0.8278
N
4,319
4,319
β1
β2
Panel B 2008 – 2009 Period Parameter Estimates (t-statistics in parentheses)
SG&A
COGS
β0
0.0118**
0.0106***
(3.24)
(3.60)
0.4792***
0.8550***
(39.36)
(86.17)
-0.0339
0.1251***
(1.52)
(6.89)
Adjusted R2
0.4439
0.7244
N
6,727
6,727
β1
β2
***, **, * statistically significant at the 1%, 5%, 10% levels for a two-tailed test.
When β2 is compared across the two periods, it is significantly lower in the SGA regression (t=4.54) and
Cost behavior of selling, Page 8
Research in Business and Economics Journal, Volume 10 – October, 2014
significantly higher in the COGS regression (t=3.23) in the 2008 – 2009 period.
Cost behavior of selling, Page 9